QWE AI Academy Two Ways to Edit Video with AI – One Wastes Money
You want to remove a person from a clip. Two approaches: generate an entirely new video from scratch, or edit the existing one.
Most beginners pick generation. Type a prompt, burn 50+ credits across 10 attempts, get one usable result. The math stings: Gen-4 costs roughly 25 credits per second, so a 5-second clip runs you 125 credits. That’s 20% of the Standard plan’s monthly allowance for one clip.
The better route? Aleph, Runway’s video transformation tool that edits existing footage by adding, removing, or changing elements. Upload your clip, prompt “remove person in red jacket,” done. Same result, fraction of the cost, because you’re manipulating real footage instead of asking the AI to hallucinate an entire scene.
But here’s the catch nobody tells you: Aleph has quirks, Video-to-Video has a 20-second hard cap, and the credit system rounds up in ways that’ll surprise you. This guide covers the actual workflow – costs included.
Why Standard Tutorials Skip the Credit Math
Every Runway tutorial shows you how to generate video. Few explain when to use which tool, or what it actually costs per clip.
The official docs list features. Community posts share cool outputs. But the decision framework – “Should I use Video-to-Video or Aleph for this edit?” – is absent. So beginners default to text-to-video generation for everything, then wonder why their 625 credits vanished in two days.
Here’s what’s usually missing:
- Model cost differences – Gen-4 Turbo costs 25 credits/5 sec, Gen-3 Alpha costs 10 credits/5 sec. That’s a 2.5x difference. Most guides say “Gen-4 is better” without mentioning you’ll hit your limit 2.5x faster.
- Credit rounding traps – Video-to-Video charges per second but rounds up to the nearest 5-second increment. A 3-second clip costs the same as 5. Generate lots of short clips and you’re overpaying by 40-60%.
- Input limits – Video-to-Video maxes out at 20 seconds of input; longer videos get truncated to the first 20 seconds only. Upload a 60-second clip? You’re editing the first third, not the whole thing.
These aren’t exotic edge cases. They’re daily realities if you’re using Runway for real projects.
The Right Approach: Match Tool to Task
Runway isn’t one tool – it’s three workflows with different cost profiles. Pick wrong and you’ll burn credits. Pick right and you’ll stretch your monthly allowance 3-5x further.
Workflow 1: Edit Existing Footage (Aleph)
Use this when you have footage that’s almost perfect. Remove an object, change lighting, swap backgrounds, add elements.
Navigate to Generative Session in your dashboard. Click Chat Mode and upload a video, or switch to Tool Mode and ensure Gen-4 is selected before uploading to the prompt canvas. Key detail: Aleph defaults to using the first 5 seconds of your video. For longer clips, use the trimming interface to select which portion you want to edit.
Write a simple prompt with an action verb: “remove the car,” “change sky to sunset,” “add fog.” Effective prompts include an action verb (add, remove, change, replace, re-light, re-style) plus a description of what changes.
Aleph supports a max duration of 5 seconds per generation. If you need to edit a longer sequence, break it into 5-second chunks. Once complete, you can upscale the result to 4K in Chat Mode.
Pro tip:For elements that must stay consistent across edits, include them in your input video. Generate consistent subjects in References first, then animate with Gen-4. This prevents the AI from changing things you want unchanged.
Workflow 2: Restyle or Transform Video (Video-to-Video)
Use this when the motion is right but the style is wrong. Turn live-action into animation, apply artistic styles, or change the look entirely while preserving the original movement.
Go to Generative Session, select Gen-3 Alpha or Gen-3 Alpha Turbo from the bottom left dropdown, then upload a video or drag-and-drop. You’ll see Video-to-Video controls appear.
Important: The output matches your input length if it’s 20 seconds or less. Longer videos? Only the first 20 seconds get processed. If you uploaded a 45-second clip expecting a 45-second stylized output, you’ll be disappointed.
Adjust the Structure slider. Lower values maintain the structure of the original video. Higher values produce more abstract outputs. For realistic restyling (turn footage into watercolor painting), keep it low. For surreal transformations, crank it up.
Write your prompt or upload a style reference image. Detailed prompts yield results more closely aligned with your concept, though simple prompts can work well too.
Cost reality check: Credits charge per second but round up to the nearest 5-second increment. That 8-second clip? You’re paying for 10. This is where batch editing gets expensive fast.
Workflow 3: Generate from Scratch (Text-to-Video)
Use this when you have no footage and need to create a clip from a text description alone. Most expensive option – reserve for shots you can’t film or edit.
Navigate to Generative Session and select Gen-3 Alpha from the dropdown (Gen-3 Alpha Turbo requires an input image). Text-only prompting is only available on the full Alpha model, not Turbo.
Add a descriptive prompt that conveys camera angle, subject, scene, style, and movement. Good structure: “[camera movement]: [establishing scene]. [additional details].” Example: “Low angle tracking shot: a woman in orange walks through a rainforest. Overcast gray sky. Vibrant flora.”
Choose between 5 or 10 second duration with the dropdown near the Generate button. Longer = more credits. A 10-second Gen-4 clip can cost 50 credits. Do that 12 times and your Standard plan is gone.
This workflow makes sense for B-roll you can’t shoot (astronaut on Mars, historical reenactment, fantasy creatures). For everything else, start with real footage and use Aleph or Video-to-Video.
Real Scenario: Editing a Product Video for $0.48
You shot a product demo. Great footage, but there’s a distracting logo on the wall behind the product. Removing it in After Effects? 20 minutes of rotoscoping per second of video.
In Runway: upload the clip to Aleph, prompt “remove wall logo,” generate. 5 seconds of footage processed. Aleph maxes at 5 seconds per generation, so if your shot is 12 seconds, you’ll run it 2-3 times on different portions, then stitch in a traditional editor.
Cost breakdown (assuming Standard plan at $12/month with 625 credits): Aleph runs on Gen-4, roughly 25 credits per 5 seconds. That’s $0.48 per 5-second edit (12/625*25). Two edits for a 12-second clip = $0.96 total. Compare that to the $50-100 you’d pay a freelancer for the same rotoscoping work.
The alternative – generating the entire product shot from scratch via text-to-video – would cost 25 credits per 5 seconds and require 15-20 generations to get the product looking accurate. You’d spend 375-500 credits ($7.20-$9.60) and still need manual compositing. Editing existing footage wins every time for this use case.
Three Settings That Actually Matter
| Setting | What It Does | When to Adjust |
|---|---|---|
| Structure (Video-to-Video) | Controls how closely output matches input structure. Low = faithful, high = abstract | Set low for realistic restyling, high for surreal transformations |
| Duration (all models) | Choose 5 or 10 seconds for text-to-video outputs | Start with 5-second tests. Only go 10 seconds once you’ve dialed in the prompt |
| Model Selection | Gen-3 Alpha Turbo = faster, Gen-3 Alpha = highest fidelity | Use Turbo for drafts and iterations. Use Alpha for final outputs only |
Everything else – aspect ratio, camera controls, keyframes – is secondary until you’ve nailed these three. Most credit waste happens because people generate 10-second clips on Gen-4 when a 5-second Turbo draft would’ve shown the prompt wasn’t working.
What Nobody Mentions About the Unlimited Plan
The Unlimited plan costs $76/month and promises unlimited video generations. Sounds perfect for heavy users. But there’s a detail buried in the docs.
Unlimited Plan generations run in Explore Mode, which means slower start times during peak periods and limits on how many videos you can create simultaneously. Translation: you’re in a queue. During busy hours, your “unlimited” generations might take 10-15 minutes to start, and you can only run 2-3 at once.
Worse: community reports suggest heavy usage triggers account reviews. Reddit users report bans for ‘suspicious activity’ after just 15 days of heavy generation on Unlimited plans. Runway hasn’t published official usage caps, but the pattern is real.
For most users, the Pro plan ($28/month with 2,250 credits) is the better deal. You get priority rendering, no queue weirdness, and predictable costs. Only go Unlimited if you’re generating 50+ clips per day and can tolerate slower processing during peak hours.
Browser and Export Gotchas
Two things will break your workflow if you don’t catch them early.
First: Runway is designed for Chromium-based browsers (Chrome, Edge, Brave). Many common issues are instantly resolved by switching from Firefox or Safari. If uploads stall or tools freeze, try Chrome before troubleshooting anything else.
Second: The video editor expects all portions of your timeline to contain visual media. Leaving gaps on the default black background will cause export failures. Add a black PNG to fill gaps or the export halts. This trips up everyone who’s used to traditional NLEs where empty timeline space is fine.
Also: Special characters (accents, unicode symbols, font characters) in filenames, text layers, or project names cause export failures. Stick to standard ASCII characters. Name your project “Product_Video_v2,” not “Product Video ✨ Final.”
FAQ
How much does a 10-second video actually cost in Runway ML?
Depends on the model. Gen-4 Turbo runs about 50 credits for 10 seconds (25 per 5 sec). Gen-3 Alpha costs 20 credits for the same duration. On the Standard plan (625 credits), that’s 12 Gen-4 clips or 31 Gen-3 clips per month. Text-to-video generation typically requires 5-10 attempts to get a usable result, so factor that into your math.
Can I edit videos longer than 20 seconds with Video-to-Video?
Not in one pass. Video-to-Video processes up to 20 seconds of input. Longer videos get truncated to the first 20 seconds automatically. Workaround: split your video into 20-second chunks, process each separately, then reassemble in a traditional editor. Or use Aleph for spot edits on specific 5-second portions – often faster than processing the whole thing.
Is the free plan enough to learn Runway ML?
The free plan gives you 125 one-time credits (not monthly). That’s enough for a handful of short test clips, but exports are watermarked and you’re limited to three projects at once. It’s sufficient to understand the interface and test prompts, but not for any real project work. Once those 125 credits are gone, you’ll need to upgrade. Best use: spend them on 5-second Turbo drafts to learn which prompts work before committing to a paid plan.
Start with the official Runway help docs to explore all tools, or dive into Gen-3 Alpha’s research page for prompt examples. If you’re building production workflows, check the API pricing guide for programmatic access costs.